Download spatial habitat heterogeneity influences competition

Ecology, 84(11), 2003, pp. 2843–2855
q 2003 by the Ecological Society of America
SPATIAL HABITAT HETEROGENEITY INFLUENCES COMPETITION AND
COEXISTENCE IN AN AFRICAN ACACIA ANT GUILD
TODD M. PALMER1
Center for Population Biology, University of California, Davis, California 95616 USA and Mpala Research Centre,
Box 555, Nanyuki, Kenya
Abstract. Spatial mosaics in resource productivity may facilitate competitive coexistence when species differ in their ability to exploit resource-rich vs. resource-poor conditions. In this study, I investigated the influence of a termite-generated spatial mosaic in
resource productivity on the dynamics of competition and coexistence in a guild of four
acacia ants that reside on Acacia drepanolobium. Near termite mounds, both new shoot
production of A. drepanolobium and densities of litter-dwelling invertebrates (an important
food source of the acacia ants) are higher than in ‘‘matrix’’ habitats between mounds. This
resource variation is spatially correlated with the outcome of competition among acacia
ants for host trees; near the mounds, competitively dominant species are more likely to
supplant subordinates, whereas the likelihood of subordinates replacing dominants on host
trees increases with distance from mounds. Variation in competitive outcomes among patch
types appears to result from differential responses among acacia ant species to resource
heterogeneity. Analysis of average tissue d15N levels indicated that acacia ant colonies near
termite mounds had higher ratios of animal prey in their diets than colonies in matrix areas.
Bait trials and analysis of average nest size in mound and matrix habitats suggest that
competitively dominant species are disproportionately successful in exploiting higher prey
densities near termite mounds. Increased dietary uptake of nitrogen-rich prey may fuel
more rapid colony growth in competitively dominant acacia ants, resulting in more pronounced asymmetries in colony size between dominant and subordinate species in mound
areas. Because colony size asymmetry is a key determinant of interspecific competitive
outcomes, colonies of subordinate species near termite mounds may be more vulnerable
to aggressive take-overs by dominants. In lower-productivity matrix habitats, increased
tolerance to low-resource conditions may afford subordinate species greater persistence.
Overall, these results suggest that termite-induced habitat heterogeneity plays a significant
role in the dynamics of the acacia ant community, and may contribute to species coexistence
in this intensely competitive community.
Key words: acacia ants, Acacia drepanolobium; ant–plant mutualism; colony size; Crematogaster;
Kenya; patch dynamics; resource competition; spatial heterogeneity; species coexistence; stable isotope; Tetraponera.
INTRODUCTION
Spatial heterogeneity is a ubiquitous feature of natural ecosystems. Classic experiments over five decades
ago (e.g., Gause 1932, Park 1948) convincingly demonstrated that this patchiness is crucial to the coexistence of competing species, and ecologists in the ensuing decades have done much to elaborate how and
when spatial variability in habitats may promote the
maintenance of species diversity (reviewed in Tilman
and Kareiva 1997, Tokeshi 1999).
Patchiness in nature takes many forms. At the landscape level, habitat patches may be relatively uniform,
or may differ from one another either qualitatively
(e.g., wooded areas vs. grassland) or quantitatively
(e.g., varying levels of productivity in a single habitat
type). Different patch characteristics lead to differing
Manuscript received 26 August 2002; revised 22 February
2003; accepted 26 February 2003. Corresponding Editor: N. J.
Gotelli.
1 E-mail: [email protected]
expectations of the mechanisms that may promote species coexistence (Hanski 1995). In relatively uniform
habitats, competing species may coexist through intraspecific aggregation (e.g., Hanski and Cambefort 1991,
Ives 1991), trade-offs in colonization and competitive
ability (e.g., Levins and Culver 1971, Tilman 1994, but
see Yu and Wilson 2001), or via priority effects (Wilbur
and Alford 1985) that lead to alternative local equilibria
(e.g., Levin 1974, Barkai and McQuaid 1988). At the
other extreme, patches may differ qualitatively from
one another, and variation in habitat selection among
species can reduce the probability of competitive exclusion (e.g., Schoener 1974).
Structurally homogeneous habitats may consist of a
mosaic of patches that vary with respect to quantitative
factors such as resource level or productivity. In this
case, spatial variation in resources may facilitate coexistence if there are interspecific trade-offs in the ability of species to exploit resource-rich vs. resource-poor
patches (Kotler and Brown 1988, Hanski 1989). For
example, different species may be superior competitors
2843
2844
TODD M. PALMER
at different resource levels (Jitts et al. 1964, Tilman et
al. 1981), or species may differ in their tolerance to
low resource conditions (Grime 1979). Inverse correlations among species in the ability to exploit rich vs.
poor resource conditions have been reported for a variety of taxa (algae, Sommer 1989; invertebrates, Kohler 1992, Schmitt 1996, Amarasekare 2000, Tessier et
al. 2000; vertebrates, Brown et al. 1994), and may play
an important role in supporting species coexistence in
mosaic habitats.
Interspecific trade-offs in competitive ability play a
key role in the organization of ant communities (Hölldobler and Wilson 1990). For example, trade-offs in
the ability to locate resources and the ability to dominate resources once located (Fellers 1987) often lead
to a temporal succession of ant species at baits in experimental studies (e.g., Fellers 1987, Savolainen and
Vepsäläinen 1988, Morrison 1996, Holway 1999, and
reviewed in Hölldobler and Wilson 1990). While temporal partitioning of resources among ants is well documented (reviewed in Cerdá et al. 1998), fewer studies
have documented the influence of spatial variation in
resource abundance on ant coexistence and community
structure (Davidson 1977, Kaspari 1996, Schooley et
al. 2000). Nonetheless, the apparent generality of competitive trade-offs among ants (Davidson 1998) suggests that spatial variation in resource levels may play
an important role in competitive coexistence in some
ant communities.
In this study, I examined the influence of spatial
variation in resource availability on patterns of competition and coexistence in a widespread guild of acacia
ants residing on Acacia drepanolobium in Laikipia,
Kenya. In this guild, four ant species coexist despite a
strong dominance hierarchy (Crematogaster sjostedti
. C. mimosae . C. nigriceps . Tetraponera penzigi)
when colonies compete for limiting host trees (Stanton
et al. 1999, Palmer et al. 2000). These ant species are
generally mutually exclusive (only one ant species is
found on a given tree), and competition for trees is
intense. More than 99% of trees .1.0 m tall are occupied by ants (Young et al. 1997), interspecific turnover of ants on host trees can be as high as 7% in a
six-month period, and violent interspecific take-overs
of host trees by adjacent colonies are common (Palmer
et al. 2000). Because these four acacia ant species exploit the same species of host tree, and the three Crematogaster species exploit the same food resources and
forage actively during similar hours (A. E. Evans, unpublished manuscript, and see Study System and Methods), opportunities for conventional resource partitioning (e.g., via niche displacement) are lessened. However, the ecosystem is characterized by strong spatial
heterogeneity in resource abundance, and this feature
may contribute to species coexistence if acacia ant species differ in their ability to exploit resource-poor vs.
resource-rich patches.
Ecology, Vol. 84, No. 11
Throughout the A. drepanolobium habitat at our
study site, there are strong gradients in the productivity
of resources upon which the acacia ants depend. These
resource gradients are highly spatially correlated with
mounds of the subterranean termite Odontotermes spp.
Through their activities (e.g., concentration of organic
materials and coarse soil particles), these termites locally enrich soils in nitrogen, phosphorus, and organic
carbon, and increase soil drainage and water availability at termite mounds (T. Palmer, unpublished data).
Relative to the surrounding matrix areas, production of
nectary-bearing new shoots on A. drepanolobium is
60% higher on host trees near termite mounds (,10 m
from mound edges). In addition, areas near termite
mounds support 2.5 times higher densities of litterdwelling invertebrates, a common food of the three
Crematogaster species (T. Palmer, unpublished data).
Consequently, termite activity in this habitat appears
to generate a spatial mosaic of relatively more productive mound areas (;21% of the habitat; T. Palmer,
unpublished data) embedded within less productive
matrix areas.
This spatial variation in productivity caused by termites may strongly influence species coexistence in this
intensely competitive community. In a prior study,
Palmer et al. (2000) demonstrated that variation in A.
drepanolobium growth rate was correlated with the outcome of competition among acacia ants competing for
limiting host trees; more subordinate species were replaced by dominants on faster growing trees, while
more dominant species tended to be replaced by subordinates on slower growing trees. However, whether
variation in host tree growth rate is a cause or correlate
of competitive outcomes was not determined.
In this study, I addressed several questions to determine how spatial variation in habitat productivity
may influence the dynamics of competition and coexistence among the acacia ants: (1) Do patterns of
host tree occupancy differ among acacia ant species in
mound vs. matrix areas? (2) Do the dynamics of competition (e.g., direction of competitive outcomes) and/
or the intensity of competition (e.g., risk of interspecific
take-over on any given tree) differ for the acacia ants
in mound vs. matrix areas? (3) Do interspecific asymmetries in colony size, the critical determinant of competitive outcomes (T. Palmer, unpublished manuscript),
vary between mound and matrix habitats? (4) Can interspecific differences among acacia ants in the ability
to exploit high productivity areas explain spatial variation in competitive outcomes? More specifically, do
foraging trials, stable 15N isotope analysis, and a resource addition experiment indicate differential responses of acacia ant species to resource heterogeneity?
STUDY SYSTEM
AND
METHODS
This research was conducted in the semi-arid Laikipia ecosystem (378 E, 08 N; 1800 m elevation) in
north-central Kenya, at the Mpala Research Centre. Ap-
November 2003
PRODUCTIVITY MOSAICS AND ANT COMPETITION
2845
proximately 43% of the Laikipia ecosystem (Taiti
1992), and much of upland East Africa, is underlain
with poorly drained ‘‘black cotton’’ vertisol soils that
support wooded grassland. In these habitats, A. drepanolobium forms a virtual monoculture in the overstory, accounting for over 97% of canopy cover (Young
et al. 1997).
Three of the four acacia ant species (C. mimosae, C.
nigriceps, T. penzigi) that reside on A. drepanolobium
are obligate associates of this tree species. The fourth,
C. sjostedti, is also found on A. seyal, which is rare at
this study site. Tetraponera penzigi, C. mimosae, and
C. nigriceps all rely on swollen thorns produced by the
tree for nesting space, while C. sjostedti generally nests
in hollowed-out cavities within the tree’s twigs and
stem. Colonies of the three Crematogaster species typically occupy multiple trees, while each T. penzigi colony usually controls only a single contiguous canopy.
With the exception of T. penzigi, all of these acacia
ants consume virtually the same food resources and
forage actively during similar hours (A. E. Evans, unpublished data). Nectar is harvested from extra-floral
nectaries present on the leaflets of host trees, and invertebrates dwelling in the litter and soil around host
trees are scavenged and occasionally preyed upon by
the acacia ant workers (Hocking 1970, T. Palmer, personal observation). Both C. sjostedti and C. mimosae
tend scale insects on host trees (Young et al. 1997). In
contrast to the Crematogaster species, Tetraponera
penzigi destroys the nectaries on its host trees (Young
et al. 1997, Palmer et al. 2002), and does not appear
to forage off host trees. Instead, this species gleans
small food particles (e.g., pollen, fungal spores) from
the surfaces of host trees, a trait common to other pseudomyrmecines (Wheeler and Bailey 1920).
boundaries appear to indicate equally abrupt changes
in underlying soil characteristics; across this boundary
between dominance by P. stramineum and P. mezianum, soil physical and chemical characteristics change
dramatically. Mound soils are higher in silt and sand
than the surrounding heavy clay vertisols, and therefore
are noticeably different in texture, and lack the characteristic deep cracks present in surrounding vertisols
(T. Palmer, unpublished data). In addition, mound soils
have markedly higher levels of organic carbon, phosphorus, and nitrogen relative to the surrounding soils
(T. Palmer, unpublished data).
Determining the presence of termite mounds,
and delineating their boundaries
Examining spatial variation in the dynamics of
competition among acacia ants
To assess whether interspecific competitive outcomes vary spatially with respect to proximity to termite mounds, I compared the distance from termite
mound edges of trees that had undergone transitions in
ant occupancy in the direction of the competitive hierarchy (i.e., dominant species replaced more subordinate species) with trees where transitions had occurred against the competitive hierarchy (i.e., subordinate species replaced more dominant species). I identified 80 trees that had undergone transitions in ant
occupancy (61 in the direction of the competitive hierarchy and 19 against the competitive hierarchy) between July 1998 and July 1999 from a long-term monitoring transect of 1773 trees (Palmer et al. 2000). The
distance of each tree to the nearest termite mound edge
was measured using a meter tape.
Odontotermes construct circular mounds that are
generally 10–20 m in diameter, but no more than 0.5
m high (Darlington and Bagine 1999). Although they
are low in profile, they can readily be delineated from
the surrounding area through a combination of vegetation, soil, and other characteristics. At this study site,
the perennial grass Pennisetum stramineum grows
abundantly on mounds (60–80% cover), and declines
sharply .1 m from the mound edge to ,25% cover
(T. Palmer, unpublished data). Equally sharply, Pennisetum mezianum increases in abundance across a fairly short distance (1–2 m) at the mound’s edge, increasing from ,20% cover on top of mounds to .50% cover
beyond the mound boundaries. These two grass species
differ strongly in appearance and texture, making the
identification and delineation of mounds more straightforward (T. Palmer, unpublished data). I used this
strong contrast in plant species composition to delineate mound ‘‘edges’’ from the surrounding vertisol
soils. Strong changes in vegetation across mound
Measuring acacia ant species composition on
A. drepanolobium host trees along transects
intersecting termite mounds
To determine whether the composition of the acacia
ant community might be influenced by proximity to
termite mounds, I recorded ant species occupying trees
along 40-m transects originating in the center of termite
mounds. I randomly selected 15 mounds for this analysis. If any termite mound selected was within 30 m
of another mound, it was discarded and another chosen
for study. At each termite mound, I located the center
(defined as the point of intersection of two diameters
running in the N–S and E–W cardinal directions), and
ran two 4-m wide belt transects out in two randomly
chosen cardinal directions. Transects were run from the
center of mounds to a distance of 30 m outside the
mound edge. Along belt transects, I recorded the distance along the transect where A. drepanolobium trees
were encountered, the heights of the trees, and their
resident ant species. For each mound, I then calculated
the percentage of trees occupied by each ant species
adjacent to (#10 m) and distant from (.10 m) mound
edges.
Examining spatial variation in ‘‘risk’’ of host tree
take-over by more dominant competitors
To determine whether proximity to termite mounds
influences a colony’s risk of losing a host tree to take-
2846
TODD M. PALMER
Ecology, Vol. 84, No. 11
over by more dominant species, I compared the distance from termite mound edges of ‘‘transitioning to
dominant’’ and ‘‘stable’’ trees from the long-term monitoring transect. ‘‘Transitioning to dominant’’ trees
were defined as trees undergoing a change in ant occupant from a subordinate species to a more dominant
species over a 12-month period. ‘‘Stable’’ trees were
defined as trees not undergoing changes in ant occupant
over a 24-month period. The same ‘‘transitioning to
dominant’’ trees (N 5 61) used in the preceding analysis were used for this analysis. For each of these 61
trees, I randomly selected a size-matched (65 cm
height and 63 cm basal diameter), ‘‘stable’’ tree that
had been continually occupied over the past 24 months
by the original ant occupant of the transitioning tree.
For each of these trees, I measured the distance to the
edge of the nearest termite mound. Matching trees by
size is critical for this analysis, since species turnover
rate is much higher on smaller trees (Palmer et al.
2000). Because I used the same set of 61 trees in both
of the above analyses, I used a Bonferroni correction
for both analyses to minimize the probability of Type
I error.
randomly chosen focal trees were transferred in clipped
swollen thorns to all neighboring trees within 10 m
occupied by conspecifics. I then assessed whether
workers from the two trees fought (indicating that trees
were occupied by different colonies) or not. Cases
where workers from neighboring trees fought were unambiguous. When neighbors appeared not to fight, I
performed a reciprocal transfer to ensure that the trees
belonged to the same colonies. Each interaction from
the transfer of swollen thorns was observed for ;10
min. When colonies occupied more than two trees, I
performed a number of reciprocal transfers between
different pairs of trees to ensure that identification of
same-colony trees was correct. When conflicting results were obtained, all trees were re-tested until I obtained an unequivocal result. After all host trees occupied by a colony were identified, I measured the
height of each tree using a meter tape. I measured fewer
C. sjostedti colonies because their colonies often occupied .20 trees, and identifying a single colony usually took at least a full day.
Measuring spatial variability in acacia ant
colony size
Spatial variation in invertebrate prey resources associated with termite mounds may differentially influence acacia ant species if they vary in their ability to
discover and recruit to these resources off of their host
trees. To assess the relative foraging abilities of the
acacia ant species, I conducted foraging trials using
tuna baits. Only the three Crematogaster species (see
Plate 1) were studied, since T. penzigi appears to be
strictly arboreal, and has not been observed foraging
off of host trees (T. M. Palmer, M. Stanton and T.
Young, unpublished data). I randomly selected trees
1.5–2.0 m in height occupied by each of the three Crematogaster species (N 5 26, 29, and 27 focal trees for
C. sjostedti, C. mimosae and C. nigriceps, respectively)
in matrix habitats. Each focal tree was chosen from a
different colony. I then placed 5 g of tuna bait on the
ground at the base of each focal tree, at a distance of
40 cm in a westerly direction from the stem. Prior
observations indicated that these three Crematogaster
species seldom recruit to baits at distances over 1.0 m
from host trees within a 2-h observation period (A. E.
Evans, unpublished data). Hence, I positioned baits at
40 cm from trees to increase the probability that baits
would be discovered by focal colonies, while reducing
the likelihood that workers from nearby trees (generally
.2.0 m distant) would discover baits. I then revisited
baits at 30-min intervals, and recorded the number of
workers present in a 10-cm radius around each bait, in
addition to recording the presence and number of any
other ant species. After counting the number of workers
at baits, I observed the movements of foragers carrying
food items to determine whether they were returning
to the focal tree or to other nearby host trees. Trials
were discontinued after 150 min of observation. I re-
Because colony size is a major determinant of interspecific competitive outcomes among these acacia
ants (T. Palmer, unpublished manuscript), I assessed
whether proximity to termite mounds affected colony
sizes for the four acacia ant species. Focal colonies
were chosen randomly along five parallel 1-km transects, each running N–S and separated by at least 100
m. Two observers walked along each transect until a
termite mound was intersected. At each mound, I measured colony size for the first colony of each species
encountered within 10 m of the mound edge. When a
particular ant species was absent from an encountered
mound, I moved to the next species. After measuring
colony size for the different species at mounds, I ran
a perpendicular 30-m transect either due east or west
(determined by coin toss). At the end point of this
transect (30 m from the mound edge), I measured colony sizes for the first colonies of each species encountered within 10 m, provided there were no termite
mounds within 10 m of that point. When a termite
mound was encountered within 10 m of the endpoints
of perpendicular transects, I established a transect in
the opposite cardinal direction for colony size measurements. Again, when a particular species was not
encountered, I moved on to the next species.
Colony size was estimated by measuring the total
height of all trees occupied by a given colony, which
provides a reliable surrogate for relative colony size
within and among the four acacia ant species (T. Palmer,
unpublished manuscript). To determine which trees
were occupied by a single colony, I used a modification
of methods given in Hölldobler (1979). Workers from
Measurements of foraging ability of the three
Crematogaster species
November 2003
PRODUCTIVITY MOSAICS AND ANT COMPETITION
2847
PLATE 1. (Left) A column of Crematogaster mimosae workers departing from and returning to the host A. drepanolobium
tree during off-tree foraging for insect parts and prey. (Right) Odontotermes spp. mound in Acacia drepanolobium wooded
grassland. The termite mound is in the foreground, delineated by the abrupt transition from darker (more green) vegetation
to lighter vegetation (less green) in the background. Note the absence of trees on the mound, and the presence of taller trees
at the mound periphery. Photographs by Todd M. Palmer.
plenished baits as they were removed by ants to keep
the abundance of the resource approximately constant
throughout the trials. If one or more workers from the
focal tree discovered the bait, it was noted that the
colony had ‘‘discovered’’ the bait. Successful ‘‘recruitment’’ was noted when .10 workers from a focal tree
were present on experimental baits at a single time. A
colony was successful at ‘‘defending’’ baits when
workers were present on baits with other ground-dwelling ant species, and then were capable of completely
displacing these other ant species from baits.
each sample in a muffle furnace for 48 h at 608C. These
samples were then re-frozen until they were analyzed
for percentage of nitrogen and d15N using a mass spectrometer (PDZ Europa Hydra 20/20 Mass Spectrometer,
Davis, California, USA) at the isotope analysis facility
at University of California, Davis. Just prior to isotopic
analysis, ants from each sample were ground using a
mortar and pestle, pooling individuals within each colony. I used whole ants for these analyses, so d15N ratios
may reflect both tissue and gastric content isotope levels.
Measurements of d15N levels and percent nitrogen in
acacia ant body tissues
STATISTICAL ANALYSES
To estimate the importance of invertebrate prey in
the diets of these acacia ants, I analyzed ratios of stable
nitrogen isotopes in all four ant species, adjacent to
(,10 m) and distant from (.20 m) termite mound edges. Higher d15N ratios are thought to indicate a greater
ratio of prey to plant resources in the diet (Minegawa
and Wada 1984, Davidson and Patrell-Kim 1996, Vasconcelos and Davidson 2000). The same samples were
analyzed for percentage of nitrogen content, which may
reflect a species’ investment in costly, protein-rich exoskeleton (see Davidson and Patrell-Kim 1996). I randomly selected 10 colonies for each species located
,10 m from termite mound edges, and 10 colonies for
each species located at distances .20 m from termite
mound edges. Samples were collected between 0700
and 0900 hours, before colonies began intensive foraging. Using latex gloves, I disturbed a single tree in
each colony and collected ;20 workers. Workers were
placed in marked paper envelopes, and transferred to
a freezer. After killing the ants in the freezer, I dried
All statistical analyses were performed using JMP
(SAS Institute 1996). Data on the proportion of trees
occupied by each species in mound and matrix areas
were arcsine square-root transformed prior to analysis.
I pooled data from the two transects run at each mound.
MANOVA was used to determine whether overall differences in species occupancy on A. drepanolobium
existed between mound and matrix areas. I then assessed whether there were differences in the proportion
of trees occupied by each species in mound and matrix
areas using paired t tests. Data for empty trees were
analyzed separately. Significance levels for multiple t
tests were adjusted using a sequential Bonferroni correction (Rice 1989).
I used logistic regression to determine whether the
direction of competitive outcomes (transitions with and
against the competitive hierarchy) on host trees differed
as a function of distance from termite mounds. The
independent variable in this analysis was distance from
termite mounds (square-root transformed), regressed
against alternative transition outcomes (i.e., transitions
TODD M. PALMER
2848
Ecology, Vol. 84, No. 11
percentage of nitrogen in body tissues among ant species and sites (mound and matrix).
RESULTS
Acacia ant species composition along transects
intersecting termite mounds
FIG. 1. Variation in acacia ant community structure in
response to termite-induced habitat heterogeneity. Each bar
represents the mean difference (61 SE) between the percentage of trees occupied by a given species near mounds
(#10 m from mound edges) and in matrix areas (.10 m from
mound edges) for 15 transects. Percentage data were arcsine
square-root transformed for statistical analysis, and to generate this figure. Asterisks indicate significant variation (P ,
0.05) in percentage occupancy between mound and matrix
habitats for a given species. Abbreviations are: C. sjo, Crematogaster sjostedti; C. mim, C. mimosae; C. nig, C. Nigriceps; and T. pen, Tetraponera penzigi
occurring in the direction of or against the competitive
hierarchy). I also used logistic regression to determine
whether distance from termite mounds was a significant
predictor of risk of host tree take-over, where ‘‘stable’’
and ‘‘transition to dominant’’ tree states were alternative outcomes.
The relationship between proximity to termite
mounds and colony size was assessed using a two-way
ANOVA. Within each species I compared average colony size at mound vs. matrix areas using planned contrasts. Repeated-measures ANOVA was used to assess
differences in worker recruitment to baits over time
among the three Crematogaster species. ANOVA was
used to examine variability in both d15N levels and
TABLE 1.
The community composition of the acacia ants varied
spatially with respect to termite mounds (Fig. 1, Table
1). The proportion of trees occupied by the competitively dominant C. sjostedti was three times higher near
termite mounds, while the proportion of unoccupied
trees, and trees occupied by the competitively subordinate C. nigriceps, was approximately three times
higher in matrix habitats (Table 1). There were no significant differences between mound and matrix areas
in the proportion of trees occupied by C. mimosae or
T. penzigi (Table 1). Unoccupied trees were relatively
rare, but were found significantly more frequently in
matrix (6.5% of trees) than in mound (1.7% of trees)
habitats (Table 1).
Spatial variation in competitive outcomes
among acacia ants
Proximity to termite mounds was a significant predictor of competitive outcomes among the acacia ants.
Near the mounds, dominants were more likely to supplant subordinates on host trees, whereas the likelihood
of subordinates replacing dominants on host trees increased with distance from mounds (logistic regression: x2 5 20.01, df 5 1, P , 0.0001; Table 2a).
Spatial variation in risk of host tree take-over
by more dominant competitors
The risk that a colony would lose a host tree to a
more dominant species varied with proximity to termite
mounds. Colonies occupying host trees closer to termite mounds were more likely to be supplanted from
those trees by more dominant competitors than colonies
on host trees located at greater distances from mounds
(logistic regression: x2 5 23.76, df 5 1, P , 0.0001;
Table 2b).
Comparison of ant species occupancy rates on trees at mound and matrix areas.
Percentage of trees occupied
Occupant
Crematogaster sjostedti
C. mimosae
C. nigriceps
Tetraponera penzigi
Empty
Mound
30.58
48.57
4.82
13.90
1.76
(5.06)
(4.23)
(2.26)
(2.57)
(0.68)
Matrix
11.04
51.03
14.81
16.60
6.51
(3.08)
(5.07)
(3.27)
(2.09)
(1.84)
t
P
4.55
20.01
22.74
20.74
22.34
,0.001***
0.991
0.015*
0.470
0.034*
Notes: Values are means with 1 SE in parentheses. MANOVA results for overall comparisons
of occupancy data for mounds vs. matrix areas: Wilks’ lambda 5 0.55, df 5 5, P 5 0.0098.
Paired t tests were used to compare mean percentage occupancy (arcsine square-root transformed) between mound and matrix areas for each species. N 5 15 for each species at each
site.
* P , 0.05; *** P , 0.001.
PRODUCTIVITY MOSAICS AND ANT COMPETITION
November 2003
2849
TABLE 2. Relationship between distance from termite
mound edges and transition direction for ant colonies on
trees from the long-term monitoring transect, and the relationship between distance from termite mound edges and
‘‘risk’’ of interspecific takeover for ant colonies on trees
from the long-term monitoring transect.
Distance from mound
edge (m)
Competitive outcome
#5.0
.5.0
a) Transition direction
Subordinate supplants dominant
Dominant supplants subordinate
3
39
16
22
b) Transition history
No transition in past 24 months
Dominant supplants subordinate
15
39
46
22
Spatial variation in average colony size
Colonies adjacent to termite mounds were on average
twice as large (as measured by the total height of trees
occupied) than colonies located in matrix areas for all
three Crematogaster species (Table 3, Fig. 2). Colonies
of T. penzigi did not differ significantly in average size
between mound and matrix areas (Fig. 2). Increases in
average colony size near termite mounds were strongest
for C. sjostedti (120% increase) and C. mimosae (100%
increase), and weakest for C. nigriceps (45% increase)
and T. penzigi (29% decrease). Consequently, disparity
in colony size between dominant and subordinate species was stronger near termite mounds than in matrix
areas (Fig. 2).
Foraging trials for the three Crematogaster species
Workers of the three Crematogaster species differed
strongly in their abilities to locate and recruit to tuna
baits placed near their host trees. In all foraging trials
where Crematogaster workers successfully discovered
and recruited to baits, workers were observed returning
to focal trees, indicating that workers originated from
the focal trees under study. In 2.5-h bait trials, C. sjostedti located and recruited to 92% of baits (24 out of
26 baits), while C. mimosae and C. nigriceps recruited
to only 59% (17 out of 29) and 18% (5 out of 27) of
baits, respectively (x2 5 33.14, df 5 2, P , 0.0001).
Once baits were discovered, the three Crematogaster
species differed significantly in the number of workers
recruiting to baits over time (Table 4). Crematogaster
TABLE 3. Two-factor ANOVA on average acacia ant colony
size at two locations (mound and matrix).
Source
df
SS
F ratio
P
Ant species
Location
Species 3 location
Error
Total
3
1
3
157
164
5.11
1.02
0.62
8.79
18.21
35.46
18.33
3.66
,0.0001
,0.0001
0.013
,0.0001
Note: Colony size was estimated as log(total height of trees
occupied) for each colony.
FIG. 2. Responses of acacia ant colony size to termiteinduced habitat heterogeneity. Colony size is estimated by
measuring the total height of trees occupied by a given colony
(T. Palmer, unpublished manuscript). Each point represents
the mean total height of trees 61 SE occupied by colonies of
a given species. Near termite mounds, N 5 7, 33, 31, and 7
colonies for C. sjostedti, C. mimosae, C. nigriceps, and T.
penzigi, respectively. In matrix areas, N 5 4, 36, 40, and 7
colonies for C. sjostedti, C. mimosae, C. nigriceps, and T.
penzigi, respectively.
mimosae had the strongest recruitment response, followed by C. sjostedti and C. nigriceps (Fig. 3). Crematogaster sjostedti and C. mimosae were better able
to defend baits from exploitation by other grounddwelling ant species than was C. nigriceps. The former
two species displaced ground-dwelling ants in 92% (24
out of 26 cases) and 94% (17 out of 18) of cases when
they located baits, respectively, while C. nigriceps only
displaced soil-dwelling ant competitors in 36% (5 out
of 14) of cases where they located baits (x2 5 19.06,
df 5 2, P , 0.0001).
Stable N isotope ratios and percentage of nitrogen in
acacia ant body tissues
Ant d15N showed both strong interspecific and spatial
variation (Fig. 4, Table 5). Among the three CremaTABLE 4. Repeated-measures ANOVA on the number of
ants observed at experimental baits during 2.5-h foraging
trials.
Source
SS
df
F ratio
P
Ant
Time
Ant 3 time
Error
Total
1159.46
1191.04
243.813
3700.78
6476.81
2
4
8
43.23
22.21
2.27
,0.0001
,0.0001
0.02
,0.0001
Note: The number of workers observed at baits was
(square-root 1 0.5)-transformed to satisfy normality requirements.
2850
TODD M. PALMER
Ecology, Vol. 84, No. 11
TABLE 5. Two-factor ANOVA on d15N levels in acacia ant
body tissues.
Source
df
SS
F ratio
Ant species
Location
Species 3 location
Error
Total
3
1
3
69
76
59.08
42.25
10.05
117.14
229.12
11.60
24.89
1.97
1.69
P
,0.0001
,0.0001
0.12
,0.0001
onies near termite mounds than colonies located further
from mounds. Percentage of nitrogen in body tissues
varied among ant species (Fig. 5), but did not vary
between mound and matrix habitats (Table 6).
DISCUSSION
togaster species, average d15N was positively correlated
with differences in off-tree foraging ability; Crematogaster sjostedti had the highest tissue d15N, while
levels in C. nigriceps were lowest. d15N in T. penzigi
was also high. Average d15N was 33% higher for col-
Results from this study suggest that the spatial mosaic of resource productivity associated with Odontotermes mounds strongly influences the dynamics of
competition and community structure in the acacia ants
of A. drepanolobium (see Plate 1). In relatively productive termite mound microhabitats, competitively
dominant species are disproportionately successful in
displacing more subordinate species from host trees,
while subordinate species are more successful in less
productive matrix areas. This spatial variation in competitive outcomes appears to contribute to coexistence
in this intensely competitive guild. Spatial heterogeneity in competitive dominance likely plays a role in
maintaining species coexistence in other species assemblages (reviewed in Dunson and Travis 1991), and
may correlate with gradients in both abiotic (e.g., soil
acidity, Tansley 1917) and biotic factors (e.g., seed
FIG. 4. Mean d15N levels for acacia ant tissues collected
from colonies in mound and matrix habitats. Bars represent
the average of d15N values 11 SE . N 5 10 samples for each
species at each location. See Fig. 1 for species abbreviations.
FIG. 5. Mean percentage of nitrogen for acacia ant tissues
collected from colonies in mound and matrix habitats. Bars
represent the average of percentage nitrogen values 11 SE .
Because there was no significant effect of spatial location
(e.g., mound vs. matrix) on percentage of nitrogen, data from
both locations were combined for each species for this figure.
N 5 20 samples for each species at each location. See Fig.
1 for species abbreviations.
FIG. 3. The mean number of workers recruiting to tuna
baits over time for the three Crematogaster species. Each
point represents the mean number of workers at baits 61 SE
in scans at 30-min intervals. This figure represents only trials
where ants successfully located baits (i.e., at least one worker
was seen at the experimental bait). N 5 26, 18, and 14 trials
for C. sjostedti, C. mimosae, and C. nigriceps, respectively.
PRODUCTIVITY MOSAICS AND ANT COMPETITION
November 2003
TABLE 6. Two-factor ANOVA on percentage of nitrogen
levels in acacia ant body tissues.
Source
df
SS
F ratio
P
Ant species
Location
Species 3 location
Error
Total
3
1
3
32
39
0.005
0.00002
0.00007
0.002
0.007
21.88
0.39
0.33
,0.0001
0.53
0.80
,0.0001
density, Brown et al. 1994). In ant communities, most
research examining the influence of habitat heterogeneity on species coexistence has focused on niche partitioning in factors such as resources or nest sites (reviewed in Hölldobler and Wilson 1990). Productivity
has been suggested as a major structuring force in ant
communities at geographical scales (Kaspari et al.
2000), but the role of local spatial variation in resource
productivity on the dynamics of ant competition and
community structure is not well known. Davidson and
colleagues (Davidson and Fisher 1991, Davidson et al.
1991, Davidson and McKey 1993, Yu and Davidson
1997) have suggested that productivity mosaics driven
by light regimes may influence competitive dynamics
in Neotropical and Asian plant–ant assemblages, based
on observations and experiments demonstrating that
competitive dominants tend to successfully occupy
faster growing hosts and more productive habitats,
while subordinates are generally restricted to slower
growing hosts and less productive habitats (but see
Vasconcelos and Davidson 2000). This study adds to
a growing body of evidence that productivity mosaics
may play an important role in mediating species coexistence among plant–ant assemblages.
Odontotermes termite mounds in this habitat are associated with both increased shoot production by A.
drepanolobium and increased densities of the invertebrate prey of the acacia ants (Palmer et al., in prep).
New shoot production by A. drepanolobium should result in increased nectar availability on host trees, since
active nectaries are found primarily on new growth (T.
Palmer, M. Stanton, and T. Young, unpublished data).
In addition, increased host tree vigor near mounds may
fuel greater exudate production by the scale associates
of C. sjostedti and C. mimosae. d15N levels in all four
acacia ant species were positively correlated with proximity to mounds, suggesting that mound sites are rich
in the animal prey of these acacia ants. In addition, all
three Crematogaster species showed increases in average colony size near mounds. By contrast, there were
no significant differences in average colony size between mound and matrix areas for T. penzigi perhaps
because this species destroys host tree nectaries (Young
et al. 1997, Palmer et al. 2002) and does not forage off
of host trees. It is not clear why T. penzigi showed
increases in d15N near termite mounds. One possibility
is that this species may feed on very small prey (e.g.,
mites), which may also be more abundant near termite
2851
mounds. Further investigation is required to clarify this
pattern.
These results suggest that increased resource availability both on and off of host trees near termite mounds
promotes increased colony growth in the Crematogaster species. Because colony growth depends on the
availability of protein (Hölldobler and Wilson 1990,
Tobin 1995), while tree-provisioned resources (e.g., extrafloral nectar and homopteran exudates) are typically
carbohydrate rich and nitrogen and protein poor (Baker
et al. 1978, Davidson and McKey 1993, Davidson and
Patrell-Kim 1996), increased invertebrate prey densities may play a larger role in facilitating Crematogaster
spp. colony growth near termite mounds. However, excess carbohydrates provided by nectar may ‘‘fuel’’ ant
activity that can be directed towards locating, recruiting
to, and defending nitrogenous resources on the ground
(Davidson 1997).
Differences among Crematogaster species in the
ability to exploit nitrogen-rich off-tree resources appear
to underlie differential colony growth among the three
species. While all three Crematogaster species showed
significant increases in average colony size near termite
mounds relative to matrix areas, colony size increases
were of the greatest magnitude in the competitively
dominant C. sjostedti and C. mimosae These differences suggest that competitive dominants are better
able to exploit the increased prey availability near termite mounds, which may lead to more rapid colony
growth by these species in high productivity areas.
Consistent with this hypothesis, Crematogaster sjostedti and C. mimosae located, recruited to, and defended baits more effectively than did the subordinate
C. nigriceps suggesting that the former two species are
not constrained by trade-offs in exploitation and interference competitive ability at concentrated resources
similar to those represented by experimental baits (e.g.,
vertebrate and large invertebrate carcasses, carnivore
dung, termite colonies, etc.). High carbohydrate yields
from both extra-floral nectaries and scale exudates may
afford these species high tempo activity (Oster and Wilson 1978) and high dynamic density (workers/m2, Hölldobler and Wilson 1990) in off-tree foraging areas,
leading to more effective resource discovery, harvesting, and defense (Davidson 1998). In a number of other
ant communities, access to excess carbohydrates may
enable species to ‘‘break’’ the trade-off in exploitation
and interference competition, enhancing their ecological dominance (Davidson 1998).
The mechanisms underlying competitive outcomes
among these acacia ants may differ between mound
and matrix areas. Near termite mounds, high resource
availability appears to fuel more rapid colony growth
of competitively dominant species, increasing colony
size asymmetry between dominants and subordinates.
In this community, interspecific wars between adjacent
colonies were won through a process of attrition, and
competitive outcomes were primarily determined by
2852
TODD M. PALMER
asymmetries in colony size (T. Palmer, unpublished
manuscript). More rapidly growing colonies of dominant species may experience a greater degree of space
limitation on host trees (e.g., if colony growth exceeds
the production of new domatia), leading to take-overs
of adjacent trees occupied by the considerably smaller
colonies of subordinates. These results are consistent
with a model proposed by Davidson et al. (1991), who
suggested that poor competitors should be displaced
on fast-growing host plants by superior competitors
whose high energy demands could be met by more
vigorously growing trees. As an extension to their model, data from this study suggests that increased plant
vigor may be positively spatially correlated with increased productivity of nitrogen-rich food resources
that are critical for colony growth. When plant ants
forage both on and off their host trees, these enhanced
nitrogen-rich resources may further fuel the expansion
of competitively dominant species.
In lower productivity matrix areas, reversals in transition outcomes (i.e., subordinates increasingly replacing dominants on host trees) may result either from
true competitive reversals, or from subordinates colonizing trees that have been abandoned by competitive
dominants (Palmer et al. 2000). Data from this study
suggest that the latter explanation is more likely to
underlie most reversals in transition outcomes. First,
almost 60% of reverse transitions involved either C.
mimosae or C. nigriceps replacing C. sjostedti on host
trees. However, even in matrix areas, C. sjostedti colonies are ;2.5 times larger on average than C. mimosae
colonies, and almost six times larger than C. nigriceps
colonies. This large disparity in average colony size
suggests that more subordinate C. mimosae and C. nigriceps colonies are unlikely to usurp trees occupied
by C. sjostedti in direct conflict. More likely, C. sjostedti may abandon lower productivity satellite trees in
these low-resource areas, especially during times of
environmental stress. Average tissue nitrogen concentrations are highest in C. sjostedti suggesting that this
ant species may invest more nitrogen in costly, proteinrich exoskeleton (see Davidson and Patrell-Kim 1996).
As a consequence, this species may be less able to
maintain positive colony growth during stressful periods (e.g., droughts) when resource levels are low,
since the cost of producing workers is relatively high.
By contrast, the competitively subordinate C. nigriceps
had the lowest average tissue nitrogen concentrations,
suggesting that this species may require lower overall
dietary nitrogen to maintain positive colony growth.
Consequently, this species may be at an advantage in
resource-poor areas, or during prolonged periods of
environmental harshness. Consistent with this hypothesis, while C. nigriceps has the lowest rank abundance
among the acacia ants at this study site (occupying 9%
of host trees), the same species has the highest rank
abundance (53%) at a nearby site (,5 km distant)
where invertebrate densities, vegetative cover, and A.
Ecology, Vol. 84, No. 11
drepanolobium height growth are markedly lower (T.
Palmer, unpublished manuscript). Further, the proportion of unoccupied A. drepanolobium trees was significantly higher in matrix relative to mound areas.
Combined, these observations suggest that tolerance to
stress (e.g., the ability to maintain positive growth rate
under low-resource conditions) may be more important
than competition in structuring this community in unproductive matrix environments (e.g., Grime 1973,
1979). Unproductive microhabitats may provide a
‘‘competitive refuge’’ for subordinate species such as
C. nigriceps, increasing their persistence in the community. The fact that the risk of host tree take-over by
more dominant species declines with distance from termite mounds lends support to this interpretation.
A further consequence of spatially contingent competitive outcomes is that the degree of intraspecific
aggregation among the acacia ants may increase if a
single species is disproportionately successful at any
given site. Aggregation serves to increase the intensity
of intraspecific, relative to interspecific, competition
(Atkinson and Shorrocks 1981, Ives and May 1985),
which may further facilitate coexistence among these
acacia ants. Development of spatially implicit and explicit models of competition in this community is underway.
While host tree occupancy by C. sjostedti and C.
nigriceps differed significantly (in opposite directions)
between mound and matrix areas, there were no significant differences in occupancy by C. mimosae or T.
penzigi. Crematogaster mimosae is a competitively intermediate species, and in transitions on marked trees
between July 1998 and July 1999, lost nearly as many
trees to the more dominant C. sjostedti (N 5 22 trees)
near termite mounds as it gained in take-overs of the
more subordinate C. nigriceps and T. penzigi (N 5 26
trees total; T. Palmer, M. Stanton, and T. Young, unpublished data). In matrix areas, C. mimosae lost six
host trees to other species, and gained five. Consequently, this species did not increase or decrease significantly in abundance between the two subhabitat
types. The lack of significant variation in the proportion
of trees occupied by T. penzigi between mound and
matrix habitats may result from the strong priority effects exerted by this species on host trees; nectary destruction and the maintenance of small entry holes in
swollen thorn domatia by T. penzigi dramatically reduce the probability of aggressive take-over of their
host trees by competitively dominant species (Palmer
et al. 2002). Similar occupancy on host trees in mound
vs. matrix areas may result because T. penzigi does not
aggressively expand onto neighboring trees, nor succumb easily to aggressive take-over.
Despite habitat-dependent variation in competitive
outcomes, all four of the acacia ant species were found
in both mound and matrix habitats. This observation
begs the question: What maintains species diversity in
both of these locations? For example, why are host trees
November 2003
PRODUCTIVITY MOSAICS AND ANT COMPETITION
in mound areas not completely dominated by C. sjostedti colonies? The answer to this question likely includes consideration of temporal environmental variability. If the larger colonies (analogous to larger body
size) of more dominant species require higher protein
intake to maintain positive growth (see Kotler and
Brown 1988, Davidson and McKey 1993), then, during
periods of environmental harshness (e.g., droughts), the
highly polydomous (i.e., multi-tree) colonies of these
species may ‘‘contract’’ from peripheral host trees,
opening up the trees to colonization by subordinate
species with superior colonization abilities (Palmer et
al. 2000, Stanton et al. 2002). Similarly, in favorable
(e.g., high rainfall) years, colonies of dominant species
may spread outward from mounds, usurping host trees
of subordinate species within matrix habitats. Preliminary data from a long-term monitoring transect are
consistent with this hypothesis, showing increases in
abundance of C. sjostedti during the year following El
Niño events, and decreases of this species in the subsequent drought year (T. Palmer, M. Stanton, and T.
Young, unpublished data). It is interesting to note that
drought negatively affected C. sjostedti despite the fact
that this species tends scale insects, suggesting that the
trophobionts did not provide a strong buffer against
strong environmental variability. East Africa is typified
by strong spatial and temporal variation in rainfall
(McClanahan and Young 1996), which may contribute
to coexistence in this system by favoring different ant
species under different rainfall regimes. We are currently developing models to examine the influence of
strong temporal variation in environmental conditions
on long-term coexistence among these acacia ants.
In addition, ‘‘spatial mass effects’’ (Shmida and Ellner 1984) may help to maintain high diversity in mound
and matrix areas, if local diversity in either patch type
is augmented by immigration from neighboring patches
with a different equilibrium community. These effects
might be significant if the recruitment of new host trees
in mound habitats is rapid enough to allow substantial
establishment by strongly colonizing subordinate species. Further data are needed to evaluate the plausibility
of this hypothesis.
Because the spatial mosaic in acacia ant resources
is closely correlated with Odontotermes mounds, the
relative density of termite mounds in this ecosystem is
an important determinant of ant community processes
and structure. The factors that mediate termite mound
densities include competitive interactions among termite colonies and resource availability to termites.
Odontotermes mounds in black cotton habitats are generally over-dispersed, a pattern thought to result from
the maintenance of foraging territories by termite colonies (Darlington 1985). The spatial pattern of mounds
is stable, and probably persists over hundreds of years
through repeated recolonization of established nest
sites (Darlington 1985, see also Watson 1967). The
overall density of mounds in a habitat is also likely
2853
limited by habitat productivity (Darlington 1985). Habitats with higher mound densities should support a
higher proportion of competitively dominant acacia ant
species, while those with lower mound densities should
be characterized by increases in the abundance of subordinate species. This hypothesis is supported by observations that the relative abundance of C. nigriceps
on A. drepanolobium trees increases more than fivefold between our study site and a nearby site where
termite mound densities are much lower (personal observation).
The repeatable patterns between local variation in
habitat productivity and competitive outcomes among
the acacia ants observed in this study may be useful
in understanding the mechanisms that underlie variation in the composition of this acacia ant community
at geographical scales. Elsewhere in Laikipia, and more
broadly across East Africa, subsets of the same four
ant species are present on A. drepanolobium. At most
other sites surveyed (Hocking 1970, T. M. Palmer, personal observation), at least three of the four ant species
can be found, indicating that the mechanisms promoting coexistence in this ant community are widespread.
However, species composition can vary dramatically
from site to site, at both local (,3 km) and regional
scales. Geographic-scale surveys of this acacia ant
community currently underway will help to determine
whether biogeographic patterns in this arboreal ant
community are associated with large-scale variation in
habitat productivity.
Results from this study provide evidence that termite-generated environmental heterogeneity plays a
strong role in structuring the acacia ant community,
influencing patterns of interspecific competition for
nest sites and contributing to coexistence among the
four ant species. As in other African savanna communities (Dangerfield et al. 1998), termites act as ‘‘ecosystem engineers’’ (sensu Jones et al. 1994) in black
cotton savannas, dramatically altering soil physical and
chemical properties. These soil changes, in turn, have
strong ‘‘upward cascading’’ impacts on the trophic dynamics of the aboveground community. Links between
ecosystem engineers and trophic ecology are understudied (Jones et al. 1997), but potentially very important to the structure and function of ecosystems
(e.g., Estes and Duggins 1995). This study provides an
example of how habitat modification by an ecosystem
engineer influences ecological patterns (e.g., variation
in community structure, intraspecific variation in colony size) and processes (e.g., competitive outcomes)
in acacia ants at both the population and community
level. Since trees occupied by the different ant species
differ strongly in growth form (Stanton et al. 1999; T.
Palmer, M. Stanton, and T. Young, unpublished data),
suites of associated invertebrate species (M. L. Stanton,
unpublished data), leaf tannin levels (Ward and Young
2002), and patterns and rates of herbivory (T. M. Palmer, unpublished data), the determinants of acacia ant
TODD M. PALMER
2854
community structure have a substantial impact on the
broader black cotton ecosystem.
ACKNOWLEDGMENTS
This research was completed in partial fulfillment of the
requirements for the degree of Doctor of Philosophy at the
University of California, Davis. Funding was provided by the
National Science Foundation (DEB 97-26663 and DEB0089706), and by a Jastro-Shields award from U.C. Davis. I
am especially grateful to John Lemboi and Amanda Evans
for help with fieldwork, and to Truman Young and Maureen
Stanton for their advice, insights, and wonderful mentorship.
Administration and staff members at the Mpala Research Centre and Mpala Farm provided excellent logistical support.
Alison Brody, Diane Davidson, Felicia Keesing, Mikaela
Huntzinger, Steve Takata, Pete Ode, Lizzie King, and Otis
Trout listened to me babble about ants for hours and made
great suggestions. Further suggestions by Nick Gotelli, Rick
Karban, and two anonymous reviewers improved earlier versions of this paper.
LITERATURE CITED
Amarasekare, P. 2000. Coexistence of competing parasitoids
on a patchily distributed host: local vs. spatial mechanisms.
Ecology 81:1286–1296.
Atkinson, W. D., and B. Shorrocks. 1981. Competition on a
divided and ephemeral resource: a simulation model. Journal of Animal Ecology 54:507–518.
Baker, H. G., P. A. Opler, and I. Baker. 1978. A comparison
of the amino acid complements of floral and extrafloral
nectars. Botanical Gazette 139:322–332.
Barkai, A., and C. McQuaid. 1988. Predator–prey reversal
in a marine benthic ecosystem. Science 242:62–64.
Brown, J. S., B. P. Kotler, and W. A. Mitchell. 1994. Foraging
theory, patch use, and the structure of a Negev Desert granivore community. Ecology 75:2286–2300.
Cerdá, X., J. Retana, and A. Manzaneda. 1998. The role of
competition by dominants and temperature in the foraging
of subordinate species in Mediterranean ant communities.
Oecologia 117:404–412.
Dangerfield, J. M., T. S. McCarthy, and W. N. Ellery. 1998.
The mound-building termite Macrotermes michaelseni as
an ecosystem engineer. Journal of Tropical Ecology 14:
507–520.
Darlingon, J. P. E. C. 1985. Lenticular soil mounds in the
Kenya highlands. Oecologia 66:116–121.
Darlington, J. P. E. C., and R. K. N. Bagine. 1999. Large
termite nests in a moundfield on the Embakasi Plain, Kenya
(Isoptera: Termitidae). Sociobiology 33:215–225.
Davidson, D. W. 1977. Species diversity and community organization in desert seed-eating ants. Ecology 58:711–724.
Davidson, D. W. 1997. The role of resource imbalances in
the evolutionary ecology of tropical arboreal ants. Biological Journal of the Linnean Society 67:153–181.
Davidson, D. W. 1998. Resource discovery versus resource
domination in ants: a functional mechanism for breaking
the trade-off. Ecological Entomology 23:484–490.
Davidson, D. W., and B. L. Fisher. 1991. Symbiosis of ants
with Cecropia as a function of light regime, Pages 289–
309 in C. R. Huxley and D. F. Cutler, editors. Ant–plant
interactions. Oxford University Press, Oxford, UK.
Davidson, D. W., R. B. Foster, R. R. Snelling, and P. W.
Lozada. 1991. Variable composition of some tropical ant–
plant symbioses. Pages 145–162 in P. W. Price, T. M. Lewinsohn, G. Wilson Fernandes, and W. W. Benson, editors.
Plant–animal interactions: evolutionary ecology in tropical
and temperate regions. John Wiley and Sons, New York,
New York, USA.
Ecology, Vol. 84, No. 11
Davidson, D. W., and D. M. McKey. 1993. The evolutionary
ecology of symbiotic ant–plant relationships. Journal of
Hymenoptera Research 2:13–83.
Davidson, D. W., and L. Patrell-Kim. 1996. Tropical arboreal
ants: why so abundant? Pages 127–140 in A. C. Gibson,
editor. Neotropical diversity and conservation. University
of California, Los Angeles (UCLA), Los Angeles Botanical
Garden, Publication Number 1, Los Angeles, California,
USA.
Dunson, W. A., and J. Travis. 1991. The roles of abiotic
factors in community organization. American Naturalist
138:1067–1091.
Estes, J. A., and D. O. Duggins. 1995. Sea otters and kelp
forests in Alaska: generality and variation in a community
ecological paradigm. Ecological Monographs 65:75–100.
Fellers, J. H. 1987. Interference and exploitation in a guild
of woodland ants. Ecology 68:1466–1478.
Gause, G. F. 1932. Experimental studies on the struggle for
existence. I. Mixed populations of two species of yeast.
Journal of Experimental Biology 9:389–402.
Grime, J. P. 1973. Competitive exclusion in herbaceous vegetation. Nature 242:344–347.
Grime, J. P. 1979. Plant strategies and vegetation processes.
John Wiley and Sons, New York, New York, USA.
Hanski, I. 1989. Population biology of Eurasian shrews: towards a synthesis. Annales Zoologica Fennici 26:339–348.
Hanski, I. 1995. Effects of landscape pattern on competitive
interactions. Pages 203–224 in L. Hansson, L. Fahrig, and
G. Merriam, editors. Mosaic landscapes and ecological processes. Chapman and Hall, London, UK.
Hanski, I., and Y. Cambefort. 1991. Dung beetle ecology.
Princeton University Press, Princeton, New Jersey, USA.
Hocking, B. 1970. Insect associations with the swollen thorn
Acacia. Transactions of the Royal Entomological Society
of London 122:211–255.
Hölldobler, B. 1979. Territories of the African weaver ant
(Oecophylla longinoda [Latreille]): a field study. Zeitschrift
fur Tierpsychologie 51:201–213.
Hölldobler, B., and E. O. Wilson. 1990. The ants. Belknap,
Cambridge.
Holway, D. A. 1999. Competitive mechanisms underlying
the displacement of native ants by the invasive Argentine
ant. Ecology 80:238–251.
Ives, A. R. 1991. Aggregation and coexistence in a carrion
fly community. Ecological Monographs 61:75–94.
Ives, A. R., and R. M. May. 1985. Competition within and
between species in a patchy environment: relations between
microscopic and macroscopic models. Journal of Theoretical Biology 115:65–92.
Jitts, H. R., C. D. McAllister, K. Stephens, and J. D. Strickland. 1964. The cell division rates of some marine phytoplankters as a function of light and temperature. Journal
of the Fisheries Research Board of Canada 21:139–157.
Jones, C. G., J. H. Lawton, and M. Shachak. 1994. Organisms
as ecosystem engineers. Oikos 69:373–386.
Jones, C. G., J. H. Lawton, and M. Shachak. 1997. Positive
and negative effects of organisms as physical ecosystem
engineers. Ecology 78:1946–1957.
Kaspari, M. 1996. Testing resource-based models of patchiness in four Neotropical litter ant assemblages. Oikos 76:
443–454.
Kaspari, M., S. O’Donnell, and J. R. Kercher. 2000. Energy,
density, and constraints to species richness: ant assemblages along a productivity gradient. American Naturalist
155:280–293.
Kohler, S. L. 1992. Competition and the structure of a benthic
stream community. Ecological Monographs 62:165–188.
Kotler, B. J., and J. S. Brown. 1988. Environmental heterogeneity and the coexistence of desert rodents. Annual Review of Ecology and Systematics 19:281–308.
November 2003
PRODUCTIVITY MOSAICS AND ANT COMPETITION
Levin, S. A. 1974. Dispersion and population interactions.
American Naturalist 108:207–228.
Levins, R., and D. Culver. 1971. Regional coexistence of
species and competition between rare species. Proceedings
of the National Academy of Sciences 6:1246–1248.
McClanahan, T. R., and T. P. Young. 1996. East African ecosystems and their conservation. Oxford University Press,
Oxford, UK.
Minegawa, M., and E. Wada. 1984. Stepwise enrichment of
15N along food chains: further evidence and the relation
between 15N and animal age. Geochimica et Cosmochimica
Acta 48:1135–1140.
Morrison, L. W. 1996. Community organization in a recently
assembled fauna: the case of Polynesian ants. Oecologia
107:243–256.
Oster, G. F., and E. O. Wilson. 1978. Caste and ecology in
the social insects. Princeton University Press, Princeton,
New Jersey, USA.
Palmer, T. M., T. P. Young, and M. L. Stanton. 2002. Burning
bridges: priority effects and the persistence of a competitively subordinate acacia ant in Laikipia, Kenya. Oecologia
133:372–379.
Palmer, T. M., T. P. Young, M. L. Stanton, and E. Wenk. 2000.
Short-term dynamics of an acacia ant community in Laikipia, Kenya. Oecologia 123:425–435.
Park, T. 1948. Experimental studies of interspecies competition. I. Competition between populations of the flour beetles, Tribolium confusum Duvall and Tribolium castaneum
Herbst. Ecological Monographs 18:267–307.
Rice, W. R. 1989. Analyzing tables of statistical tests. American Naturalist 43:223–225.
SAS Institute. 1996. JMP in statistics. Cary, North Carolina,
USA.
Savolainen, R., and K. Vepsäläinen. 1988. A competition
hierarchy among boreal ants: impact on resource partitioning and community structure. Oikos 51:135–155.
Schmitt, R. J. 1996. Exploitative competition in mobile grazers: trade-offs in use of a limited resource. Ecology 77:
408–425.
Schoener, T. W. 1974. Resource partitioning in ecological
communities. Science 185:27–39.
Schooley, R. L., B. T. Bestelmeyer, and J. F. Kelly. 2000.
Influence of small-scale disturbances by kangaroo rats on
Chihuahuan Desert ants. Oecologia 125:142–149.
Shmida, A., and S. Ellner. 1984. Coexistence of plant-species
with similar niches. Vegetatio 58:29–55.
Sommer, U., editor. 1989. Plankton ecology: succession in
plankton communities. Springer-Verlag, Berlin, Germany.
Stanton, M. L., T. M. Palmer, and T. P. Young. 2002. Competition–colonization trade-offs in a guild of African acacia
ants. Ecological Monographs 72:347–363.
Stanton, M. L., T. M. Palmer, T. P. Young, M. L. Turner, and
A. Evans. 1999. Sterilization and canopy modification of
2855
a swollen thorn acacia tree by a plant-ant. Nature 401:578–
581.
Taiti, S. W. 1992. The vegetation of Laikipia District, Kenya.
Laikipia-Mt. Kenya Papers B-2. University of Nairobi,
Kenya, and University of Bern, Switzerland.
Tansley, A. G. 1917. On competition between Galium saxatile L. (G. hercynicum) and Galium sylvestre Poll. (G.
asperum Schreb.) on different types of soil. Journal of Ecology 5:173–179.
Tessier, A. J., M. A. Leibold, and J. Tsao. 2000. A fundamental trade-off in resource exploitation by Daphnia and
consequences to plankton communities. Ecology 81:826–
841.
Tilman, D. 1994. Competition and biodiversity in spatially
structured habitats. Ecology 75:2–16.
Tilman, D., and P. Kareiva. 1997. Spatial ecology: the role
of space in population dynamics and interspecific interactions. Princeton University Press, Princeton, New Jersey,
USA.
Tilman, D., M. Matson, and S. Langer. 1981. Competition
and nutrient kinetics along a temperature gradient: an experimental test of a mechanistic approach to niche theory.
Limnology and Oceanography 26:1020–1033.
Tobin, J. E. 1995. Ecology and diversity of tropical forest
canopy ants. Pages 129–146 in M. D. Lowman and N. M.
Nadkarni, editors. Forest canopies. Academic Press, New
York, New York, USA.
Tokeshi, M. 1999. Species coexistence: ecological and evolutionary perspectives. Blackwell Scientific, Oxford, UK.
Vasconcelos, H. L., and D. W. Davidson. 2000. Relationship
between plant size and ant associates in two Amazonian
ant-plants. Biotropica 32:100–111.
Ward, D., and T. P. Young. 2002. Effects of large mammalian
herbivores and ant symbionts on condensed tannins of Acacia drepanolobium in Kenya. Journal of Chemical Ecology
28:913–929.
Watson, J. P. 1967. A termite mound in an iron-age burial
ground in Rhodesia. Journal of Ecology 55:663–669.
Wheeler, W. M., and I. W. Bailey. 1920. The feeding habits
of pseudomyrmecine and other ants. Transactions of the
American Philosophical Society 22:235–279.
Wilbur, H. W., and R. A. Alford. 1985. Priority effects in
experimental pond communities: responses of Hyla to Bufo
and Rana. Ecology 66:1106–1114.
Young, T. P., C. H. Stubblefield, and L. A. Isbell. 1997. Ants
on swollen-thorn acacias: species coexistence in a simple
system. Oecologia 109:98–107.
Yu, D. W., and D. W. Davidson. 1997. Experimental studies
of species-specificity in Cecropia–ant relationships. Ecological Monographs 67:273–294.
Yu, D. W., and H. B. Wilson. 2001. The competition–colonization trade-off is dead; long live the colonization-competition trade-off. American Naturalist 158:49–63.